Text of an entry to appear in The Cambridge Encyclopedia of Language Sciences, edited

by Patrick Hogan. Cambridge University Press.

PRINCIPLES AND PARAMETERS THEORY AND LANGUAGE ACQUISITION

William Snyder and Diane Lillo-Martin

University of Connecticut

Nativism
The basic idea in PRINCIPLES AND PARAMETERS THEORY is to distinguish
the invariants of human language (the principles) from the major points of cross-
linguistic variation (the parameters). Both principles and parameters are taken to reflect
innately determined, biological characteristics of the human brain (see UNIVERSAL
GRAMMAR). In the course of normal child development, however, the two diverge: The
principles come to operate in much the same way in every child, with minimal sensitivity
to the childs environment, while the parameters take on distinct values as a function of
the childs linguistic input.
The term parameter is normally reserved for points of narrowly restricted
variation. The Principles and Parameters (P&P) framework also acknowledges that
languages vary in ways that are relatively unconstrained by Universal Grammar, such as
the exact form of vocabulary items. These latter points of variation are usually treated as
arbitrary idiosyncrasies, to be listed in the LEXICON.
The P&P framework has its origins in the two foundational questions of modern
linguistics (Chomsky 1981): What exactly do you know, when you know your native
language? And how did you come to know it? A satisfactory answer to these questions
must address the POVERTY OF THE STIMULUS, including the fact that children are not
reliably corrected when they make a grammatical error (Brown and Hanlon 1970; Marcus
1993).
Despite the poverty of the stimulus, by the age of about five years we observe
uniformity of success at language acquisition (Crain and Lillo-Martin 1999): Aside
from cases of medical abnormality, or isolation from natural-language input, every child
acquires a grammar that closely resembles the grammar of his or her caregivers.
Moreover, even when a child is younger, and still engaged in the process of language
acquisition, extraordinarily few of the logically possible errors are actually observed in
the childs spontaneous speech (Snyder 2007). Clearly children do not acquire grammar
through simple trial-and-error learning.
Linguists working in the P&P tradition have concluded that a great deal of
grammatical information must already be present in the childs brain at birth. Of course,
different languages of the world exhibit somewhat different grammars, but the claim in
P&P is that the options for grammatical variation are extremely limited. On the P&P
approach, the childs task during language acquisition is akin to ordering food in a
restaurant: One need only make selections from a menu, not give the chef a recipe.
In other words, the information required for the child to select an appropriate
grammar from among the options is far less, both in quantity and in quality, than would
be required to build a grammar from the ground up. First, grammars that cannot be
attained with the available parameter settings will never be hypothesized by the child,
even if they are compatible with the childs linguistic input up to that point. Second, to
the extent that parameters are abstract, and thus have wide-spread consequences, a variety
of different sentence-types in the linguistic input can help the child select the correct
option. The challenge of identifying the correct grammar is still considerable, but is far
more tractable than it would be if the child had to rely on general learning strategies
alone.

Investigating Language and Its Acquisition within a P&P Framework

The P&P framework was first clearly articulated for syntax, in the context of
Government and Binding Theory (e.g. Chomsky 1981, 1986). Yet, the framework is
considerably more general. First, the same basic architecture has been applied to
phonology, notably in the framework of Government Phonology (e.g. Kaye,
Lowenstamm, and Vergnaud 1990), and also (in certain work) to semantics and
morphology. Second, recent syntactic and phonological research in the Minimalist
Program (Chomsky 1995, 2001) and in Optimality Theory (Prince and Smolensky 2004)
still crucially assumes a P&P framework, in the broad sense that it posits universal
principles and narrowly restricted options for cross-linguistic variation. (This point will
be discussed further in the next section.)
Within the P&P framework, research on childrens acquisition of language plays a
number of important roles. First, such research can clarify the Logical Problem of
Language Acquisition, which any explanatorily adequate linguistic theory must
address: How in principle can the correct grammar be chosen from among the proposed
options, using only the types of linguistic input that children actually need for successful
language acquisition? (See DESCRIPTIVE, OBSERVATIONAL, AND EXPLANATORY
ADEQUACY.) Acquisition research can help determine which types of linguistic input
are (and are not) in fact necessary, for children to succeed at language acquisition.
For example, some of the most compelling evidence for the irrelevance of
corrective feedback comes from Eric H. Lennebergs (1967, 305-9) study of a
hypolingual child. Despite the fact that the child had been mute since birth, and therefore
had had no possibility of producing any errors to be corrected, he performed at an age-
appropriate level on comprehension tests of English grammar. Hence, receiving
corrective feedback on ones own utterances seems to be unnecessary. Hearing the
linguistic utterances of other speakers, produced in context, can suffice. To achieve
explanatory adequacy, a linguistic theory must be able to account for this.
A second role of acquisitional evidence within the P&P framework lies in testing
the acquisitional predictions of proposed linguistic principles. All else being equal, if one
proposes that a given property of language is an innate principle of Universal Grammar,
then one expects the principle to be operative in children as early as we can test for it. (A
notable exception is found in the work of Hagit Borer and Ken Wexler 1992, who
propose that several specific linguistic principles undergo maturational change during
childhood.)
For example, Stephen Crain and MineharuNakayama (1987) conducted an
acquisitional test of structure dependence, the proposed principle that syntactic
movement is always sensitive to hierarchical structure. Their study tested the prediction
that structure dependence, as an innate principle, should be operative very early. The
study was conducted with three- to five-year-old children acquiring English (who were
the youngest subjects capable of performing the experimental task), and used prompts
such as the following: Ask Jabba if [the man who is beating a donkey] is mean.
Crucially, children never produced errors of the form, Is [the man who __ beating a
donkey] is mean? Such errors might have been expected, however, if the children had
been at liberty to hypothesize structure-independent rules (such as Move the first
auxiliary to the beginning of the sentence).
Third, by proposing a parameter of Universal Grammar, one makes predictions
about the time course of child language acquisition. These predictions may involve
concurrent acquisition or ordered acquisition. To see this, suppose that two grammatical
constructions A and B are proposed to have identical pre-requisites, in terms of
parameter-settings and lexical information. A and B are then predicted to become
grammatically available to any given child concurrently, that is, at the same point
during language acquisition.
A prediction of ordered acquisition results when the proposed linguistic pre-
requisites for one construction (A) are a proper subset of the pre-requisites for another
construction (B). In this case A might become available to a given child earlier than B, if
the child first acquires the subset of Bs pre-requisites that are necessary for A.
Alternatively, A and B might become available to the child concurrently, if the last-
acquired pre-requisite for B is also a pre-requisite for A. In contrast, no child should
acquire B significantly earlier than A.
As a concrete example, consider William Snyders (2001) work on the
compounding parameter (TCP). Theoretical research had suggested a link (at least in
Dutch and Afrikaans) between the verb-particle construction (cf. Mary lifted the box up)
and morphological compounding (cf. banana box, for a box where bananas are kept).
Snyder observed a one-way implication in the data from a sizable number of languages:
If a language permits the verb-particle construction, then it also allows free creation of
novel compounds like banana box. The implication is unidirectional, however: There do
exist languages that allow this type of compounding, yet lack the verb-particle
construction. Snyder therefore proposed that the grammatical pre-requisite for the
English type of compounding (i.e., the positive setting of TCP) is one of several pre-
requisites for the verb-particle construction.
A clear acquisitional prediction followed: Any given child acquiring English will
either acquire compounding first (if [+TCP] is acquired prior to the other pre-requisites
for the verb-particle construction), or acquire compounding and the verb-particle
construction at the same time (if [+TCP] is the last-acquired pre-requisite for the verb-
particle construction). In no case will a child acquire the verb-particle construction
significantly earlier than compounding. This prediction received strong support from a
longitudinal study of ten children.
This example illustrates how the investigation of language acquisition and the
investigation of mature grammars can be mutually reinforcing activities within the P&P
framework. Another example is provided by the work of Diane Lillo-Martin and Ronice
Mller de Quadros (2005), who considered the parametric pre-requisites for the different
types of wh-questions in American Sign Language (ASL), according to two competing
syntactic analyses. The two analyses yielded distinct predictions about the time course of
acquisition, which were then successfully tested against longitudinal data from children
acquiring ASL.

Areas of debate
We will mention here two areas of debate within the P&P approach to child
language acquisition, and of course there are others. (1) What types of parameters,
exactly, is the child required to set? (2) What are the observable consequences of an
unset or mis-set parameter?
One point of disagreement in the P&P literature quite generally, including the
acquisition literature, concerns the proper conception of parameters. A classic
conception, which Noam Chomsky (1986, 146) attributes to James Higginbotham, is the
switchbox metaphor: Each parameter is like an electrical switch, with a small number of
possible settings.
Yet, this is only one of many possible ways that parameters could work. A
radically different conception is found in Optimality Theory, which posits a universal set
of violable constraints. Instead of choosing particular settings for switches in a
switchbox, the learner has to rank the constraints correctly. The result is a narrowly
restricted set of options for the target grammar, as required by the P&P framework.
(Indeed, on the mathematical equivalence of a constraint ranking to a set of switchbox-
style dominance parameters, see Tesar and Smolensky 2005, 45-46.)
Still another approach to parameters is to connect them to the lexicon. (See
LEXICAL LEARNING HYPOTHESIS.) This is conceptually attractive because the lexicon
is independently needed as a repository of information that varies across languages.
Exactly what it means to connect parameters to the lexicon, however, has been open to
interpretation.
One idea is to connect points of abstract grammatical (e.g. syntactic) variation to
the paradigms of inflectional morphology. The idea is that paradigmatic morphology has
to be stored in the lexicon anyway, and might provide a way to encode parametric
choices. This approach can be found in (Borer 1984) and (Lillo-Martin 1991), for
example. A related idea is to encode parametric choices in the morphology of closed-
class lexical items. A good example is Pierre Picas (1984) proposal to derive cross-
linguistic variation in the binding domain of a reflexive pronoun from the pronouns
morphological shape. A variant of Picas approach is to encode parametric choices as
abstract (rather than morphologically overt) properties of individual lexical items. This is
the Lexical Parameterization Hypothesis of Wexler and Rita Manzini (1987), who took
this approach to cross-linguistic variation in the binding domain for both reflexives and
pronominals.
Yet another idea is to encode cross-linguistic grammatical variation in the abstract
(often phonetically null) features of functional heads. Chomsky (1995, Chapter 2) takes
this approach to V-raising in French, for example, and its absence in English: In French,
the functional head Agr0 is strong, and causes the verb to move up and adjoin to Agr0
before the sentence is pronounced. The result is the word order in Jean [AgrP voit [VP
souvent [VP Vt Marie]]], literally John [AgrP sees [VP often [VP Vt Mary]]], in place of
English John [AgrP [VP often [VP sees Mary]]].
Chomskys approach is lexical in the sense that the morphosyntactic features of
functional heads like Agr0 are taken to be listed in the lexicon. Note, however, that the
possible features of a functional head are still assumed to be quite narrowly restricted.
Thus, where earlier work might have posited a switch-like parameter of [ Verb Raising],
for example, Chomsky instead posits a choice between a strong feature versus a weak
feature on Agr0, and assumes that this particular lexical item will be present above the VP
in most or all cases. For purposes of language acquisition, the difference is extremely
minor; the child makes a binary choice, and it has consequences across a wide range of
sentence types. Therefore Chomsky's approach still falls squarely within the P&P
framework.
The second and final point of disagreement that we will mention here concerns
the consequences of unset or mis-set parameters. For concreteness we will focus on
the switchbox model: Can a switch be placed in an intermediate, unset position?
Alternatively, must a child sometimes make temporary use of a setting that is not in fact
employed in the target language? If so, what are the consequences for the functioning of
the language faculty?
One school of thought is that there is no such thing as an unset parameter: Every
parameter is always in a determinate setting, be it an arbitrary setting (cf. Gibson and
Wexler 1994), or a pre-specified default setting (e.g. Hyams 1986). On this view,
temporary mis-settings may be routine during the period when language acquisition is
still underway. (The notion that certain parameter settings might be defaults, or
"unmarked options," has its roots in the phonological concept of MARKEDNESS.)
A second school of thought maintains that parameters are initially unset. Virginia
Valian (1991) proposes that an unset parameter permits everything that any of its
potential values would allow. Somewhat similarly, Charles D. Yang (2002) proposes that
the learner begins the language acquisition process not with a single grammar, but rather
with a multitude of different grammars, all in competition against one another. Every
grammar corresponding to a permissible array of parameter-settings is included. A
consequence is that competing values of the same parameter can be in play at the same
time.
A cross-cutting view is that children may temporarily entertain non-adult
parameter settings (whether default or not; see e.g. Thornton and Crain 1994). Children
may then produce utterances which use a grammatical structure found in some of the
worlds languages, but not in the target. On this view, what is crucial is simply that the
learner must eventually arrive at the target parameter setting, regardless of what
parameter settings have been temporarily adopted along the way. This is the learning
problem that is addressed by Edward Gibson and Wexler's (1994) Trigger Learning
Algorithm, for example.
An alternative view is that the child reserves judgement on any given parameter
setting until she has enough information to set it with confidence. Initially the parameter
is in an unset state, but this time the consequence is that none of the grammatical options
tied to a specific setting of the parameter is actually endorsed by the child. Snyder (2007)
advances this view when he argues that children who are speaking spontaneously, in a
natural setting, make astonishingly few of the logically possible grammatical errors. The
vast majority of the errors that do occur are either errors of omission, or belong to a tiny
subset of the logical possibilities for "comission" errors (where the words are actually
pronounced in configurations that are ungrammatical in the target language).
Most of the grammatical comission errors that are found in studies of elicited
production or comprehension are absent from children's spontaneous speech, even when
the opportunities exist for the child to make them. Snyder concludes that many of these
errors result from the demands of the experimental tasks. When left to their own devices,
children successfully avoid putting words together in ways that would require them to
make a premature commitment to a particular parameter setting.

Conclusion
Language acquisition is a rich source of evidence about both the principles and
the parameters of the human language faculty. For this reason, research on language
acquisition plays a central role in the P&P framework.

Works Cited and Suggestions for Further Reading:

Borer, Hagit. 1984. Parametric Syntax: Case Studies in Semitic and Romance
Languages. Dordrecht: Foris.
Brown, Roger, and Camille Hanlon. 1970. Derivational Complexity and Order of
Acquisition in Child Speech. In Cognition and the Development of Language, ed.
John R. Hayes, 155-207. New York: Wiley.
Borer, Hagit, and Ken Wexler. 1992. Bi-unique relations and the maturation of
grammatical principles. Natural Language and Linguistic Theory 10:147-189.
Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam. 1986. Knowledge of Language: Its Nature, Origin, and Use. New
York: Praeger.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, Noam. 2001. Derivation by Phase. In Ken Hale: A Life in Language, ed.
Michael Kenstowicz, 1-52. Cambridge, MA: MIT Press.
Crain, Stephen, and Diane Lillo-Martin. 1999. Linguistic Theory and Language
Acquisition. Cambridge, MA: Blackwell.
Crain, Stephen and Mineharu Nakayama. 1987. Structure Dependency in Grammar
Formation. Language 63:522543
Gibson, Edward and Kenneth Wexler. 1994. Triggers. Linguistic Inquiry 25:355-407.
Hyams, Nina. 1986. Language Acquisition and the Theory of Parameters. Dordrecht:
Reidel.
Kaye, Jonathan D., Jean Lowenstamm, and Jean-Roger Vergnaud. 1990. Constituent
Structure and Government in Phonology. Phonology 7:193-231.
Lenneberg, Eric H. 1967. Biological Foundations of Language. New York: Wiley.
Lillo-Martin, Diane C. 1991. Universal Grammar and American Sign Language: Setting
the Null Argument Parameters. Dordrecht: Kluwer.
Lillo-Martin, Diane, and Ronice Mller de Quadros. 2005. The Acquisition of Focus
Constructions in American Sign Language and Lngua de Sinais Brasileira. In
BUCLD 29 Proceedings, ed. Alejna Brugos, Manuella R. Clark-Cotton, and
Seungwan Ha, 365-375. Somerville, MA: Cascadilla Press.
Marcus, Gary F. 1993. Negative Evidence in Language Acquisition. Cognition 46:53-
85.
Pica, Pierre. 1984. On the Distinction between Argumental and Nonargumental
Anaphors. In Sentential complementation, ed. Wim de Geest and Yvan Putseys, 185-
94. Dordrecht: Foris.
Prince, Alan, and Paul Smolensky. 2004. Optimality Theory: Constraint Interaction in
Generative Grammar. Malden, MA: Blackwell.
Snyder, William. 2001. On the nature of syntactic variation: Evidence from complex
predicates and complex word-formation. Language 77:324-342.
Snyder, William. 2007. Child Language: The Parametric Approach. Oxford: Oxford
University Press.
Tesar, Bruce, and Paul Smolensky. 2000. Learnability in Optimality Theory. Cambridge,
MA: MIT Press.
Thornton, Rosalind, and Stephen Crain. 1994. Successful cyclic movement. In
Language Acquisition Studies in Generative Grammar, ed. Teun Hoekstra and
Bonnie D. Schwartz, 215-253. John Benjamins.
Valian, Virginia. 1991. Syntactic subjects in the early speech of American and Italian
children. Cognition 40:21-81.
Wexler, Kenneth and Rita Manzini. 1987. Parameters and learnability in Binding
Theory. In Parameter Setting, ed. Thomas Roeper and Edwin Williams, 41-76.
Dordrecht: Reidel.
Yang, Charles D. 2002. Knowledge and Learning in Natural Language. Oxford: Oxford
University Press.